search

microsoft / CodeXGLUE

CodeXGLUE
MIT License
1.5k stars 363 forks source link

[Code Completion] Detail of CodeGPT hyperparameters #140

Closed wannita901 closed 1 year ago

wannita901 commented 1 year ago

Hi,

I tried to find the hyperparameter values(e.g. batch size, gradient accumulation steps) used in fine-tuning CodeGPT and CodeGPT-adapt for code completion tasks.

Is it the same as in README examples? i.e., batch size=2, gradient accumulation steps=4, etc.

celbree commented 1 year ago

The hyperparameters we used for the experiments are the same as in README examples.